/*
 * dm-blkdev.c
 *
 * Copyright (C) 2001 Sistina Software
 *
 * This software is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation; either version 2, or (at
 * your option) any later version.
 *
 * This software is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with GNU CC; see the file COPYING.  If not, write to
 * the Free Software Foundation, 59 Temple Place - Suite 330,
 * Boston, MA 02111-1307, USA.
 */

#include <linux/config.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <asm/atomic.h>
#include "dm.h"

struct dm_bdev {
        struct list_head list;
        struct block_device *bdev;
        atomic_t use;
};

#define DMB_HASH_SHIFT 8
#define DMB_HASH_SIZE (1 << DMB_HASH_SHIFT)
#define DMB_HASH_MASK (DMB_HASH_SIZE - 1)

/* 
 * Lock ordering: Always get bdev_sem before bdev_lock if you need both locks.
 *
 * bdev_lock: A spinlock which protects the hash table
 * bdev_sem: A semaphore which protects blkdev_get / blkdev_put so that we
 *           are certain to hold only a single reference at any point in time.
 */
static kmem_cache_t *bdev_cachep;
struct list_head bdev_hash[DMB_HASH_SIZE];
static rwlock_t bdev_lock = RW_LOCK_UNLOCKED;
static DECLARE_MUTEX(bdev_sem);

/*
 * Subject to change... seems the best solution for now though...
 */
static inline unsigned dm_hash_bdev(struct block_device *bdev)
{
	unsigned hash = (unsigned)bdev->bd_dev;
	hash ^= (hash >> DMB_HASH_SHIFT);
	return hash & DMB_HASH_MASK;
}

static struct dm_bdev *__dm_get_device(struct block_device *bdev, unsigned hash)
{
	struct list_head *tmp, *head;
	struct dm_bdev *b;
	
	tmp = head = &bdev_hash[hash];
	for(;;) {
		tmp = tmp->next;
		if (tmp == head)
			break;
		b = list_entry(tmp, struct dm_bdev, list);
		if (b->bdev != bdev)
			continue;
		atomic_inc(&b->use);
		return b;
	}

	return NULL;
}

static struct block_device *dm_get_device(struct block_device *bdev)
{
	struct dm_bdev *d, *n;
	int rv = 0;
	unsigned hash = dm_hash_bdev(bdev);

	read_lock(&bdev_lock);
	d = __dm_get_device(bdev, hash);
	read_unlock(&bdev_lock);

	if (d)
		return d->bdev;

	n = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);
	if (!n)
		return ERR_PTR(-ENOMEM);

	n->bdev = bdev;
	atomic_set(&n->use, 1);

	down(&bdev_sem);
	read_lock(&bdev_lock);
	d = __dm_get_device(bdev, hash);
	read_unlock(&bdev_lock);

	
	if (!d) {
		rv = blkdev_get(d->bdev, FMODE_READ | FMODE_WRITE, 0, BDEV_FILE);
		if (rv == 0) {
			write_lock(&bdev_lock);
			list_add(&bdev_hash[hash], &n->list);
			d = n;
			n = NULL;
			write_unlock(&bdev_lock);
		}
	}
	if (n) {
		kmem_cache_free(bdev_cachep, n);
	}
	if (rv) {
		d = ERR_PTR(rv);
	}
	up(&bdev_sem);

	return d->bdev;
}

struct block_device *dm_blkdev_get(const char *path)
{
	struct nameidata nd;
	struct inode *inode;
	struct block_device *bdev;

	if (!path_init(path, LOOKUP_FOLLOW, &nd))
		return ERR_PTR(-EINVAL);

	if (path_walk(path, &nd))
		return ERR_PTR(-EINVAL);

	inode = nd.dentry->d_inode;
	if (!inode) {
		bdev = ERR_PTR(-ENOENT);
		goto out;
	}

	bdev = dm_get_device(inode->i_bdev);
out:
	path_release(&nd);
	return bdev;
}

static void dm_blkdev_drop(struct dm_bdev *d)
{
	down(&bdev_sem);
	write_lock(&bdev_lock);
	if (atomic_read(&d->use) == 0) {
		list_del(&d->list);
	} else {
		d = NULL;
	}
	write_unlock(&bdev_lock);
	if (d) {
		blkdev_put(d->bdev, BDEV_FILE);
		kmem_cache_free(bdev_cachep, d);
	}
	up(&bdev_sem);
}

int dm_blkdev_put(struct block_device *bdev)
{
	struct dm_bdev *d;
	int do_drop = 0;
	unsigned hash = dm_hash_bdev(bdev);

	read_lock(&bdev_lock);
	d = __dm_get_device(bdev, hash);
	if (d) {
		/* 
		 * One for ref that we want to drop,
		 * one for ref from __dm_get_device()
		 */
		if (atomic_sub_and_test(2, &d->use))
			do_drop = 1;
	}
	read_unlock(&bdev_lock);

	if (do_drop)
		dm_blkdev_drop(d);

	return (d != NULL) ? 0 : -ENOENT;
}

EXPORT_SYMBOL(dm_blkdev_get);
EXPORT_SYMBOL(dm_blkdev_put);

int dm_init_blkdev(void)
{
	int i;

	for(i = 0; i < DMB_HASH_SIZE; i++)
		INIT_LIST_HEAD(&bdev_hash[i]);

	bdev_cachep = kmem_cache_create("dm_bdev", sizeof(struct dm_bdev),
					0, 0, NULL, NULL);
	if (bdev_cachep == NULL)
		return -ENOMEM;
	return 0;
}

void dm_cleanup_blkdev(void)
{
	if (kmem_cache_destroy(bdev_cachep))
		printk(KERN_ERR "Device Mapper: dm_bdev cache not empty\n");
}